Research & Publication

RESEARCH

Google Scholar: https://scholar.google.com/citations?user=4tdW_jYAAAAJ&hl=en

ResearchGate and orchid

FIRST AUTHORED JOURNAL PUBLICATIONS - 1

SkinNet-14: A fine-tuned CCT model for classifying skin cancer addressing computational complexity and training time.

Neural Computing and Applications, Status: Published

DOI: https://doi.org/10.1007/s00521-024-10225-y

Discipline: Computer Vision, Medical Image, Deep Learning, Skin Cancer, Classification.

CO-AUTHORED JOURNAL PUBLICATIONS - 2

Fast and Efficient Lung Abnormality Identification With Explainable AI: A Comprehensive Framework for Chest CT Scan and X-Ray Images.

DOI: 10.1109/ACCESS.2024.3369900

IEEE Access. Status: Published

Discipline: Explainable AI, X-ray imaging , Lung , Computed tomography , COVID-19 , Biomedical imaging , Computational modeling , Solid modeling , Pulmonary diseases , Convolutional neural networks , Generative adversarial networks ,

An Effective Approach to Address Processing Time and Computational Complexity Employing Modified CCT for Lung Disease Classification - Click here to see the Paper

https://doi.org/10.1016/j.iswa.2022.200147

Intelligent System with Applications, Elsevier, Status: Published

Discipline: Computer Vision, Medical Image, image preprocessing, Deep Learning, Vision Transformer, GAN, chest x-ray, COVID-19, lung disease, training time, time complexity

FIRST AUTHORED CONFERENCE PUBLICATIONS - 1

A Robust Deep Learning based Framework for High-Precision Detection of Liver Disease - Click here to see the Paper

ICCCM '22 - Status: Published – July, 2022

Discipline: Medical Data, Data preprocessing, kidney disease, classification, Machine Learning, Deep Learning, CNN

Abstract

Globally, chronic liver disease is a significant cause of death, affecting a large number of people. The liver can be damaged by several factors. Obesity, undiagnosed hepatitis, and alcohol abuse, to name a few examples. This is the cause of inappropriate nerve function, blood in the cough or vomit, renal failure, liver failure, jaundice, liver encephalopathy, and many other symptoms. We use the UCI machine learning Indian Liver Patient dataset, which has 583 samples and 11 characteristics. Deep learning techniques are used to detect sickness early. Artificial Neural Network (ANN), Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM) were the three methods used in this paper. Different measurement approaches, such as accuracy, precision, recall and f-1 score, false positive rate, negative rate, mean error etc. were used to check the performance of different techniques. In terms of accuracy, CNN, ANN, LSTM were found to be 96.58%, 95.72% and 99.23% accurate in each of these categories. We also used SVM to see the effectiveness of machine learning in this prediction and our accuracy for this was 86.23%. According to the research, the LSTM had the highest accuracy. By analyzing clinical data, we also explored other ways to display this information.

Full paper

click

Under Review PUBLICATIONS - 1

Ensemble-Based Machine Learning Proposal for Enhanced PCOS Diagnosis.

Journal of Computer Information Systems, Taylor and Francis journal. Status: Under Review

Discipline: Medical Data, Data preprocessing, PCOS disease, classification, Machine Learning, Ensemble method

Abstract

Polycystic Ovarian Syndrome (PCOS) is the most common hormone condition among women of reproductive age. It affects fertility and causes significant

health risks. The causes and development of the disease are not yet fully understood although early detection of PCOS can result in effective therapy. Recent breakthroughs in machine learning demonstrate promise for medical diagnostics that may assist in the early detection of PCOS. In addition, enhancing feature selection methods can identify the most essential features, reduce computing time, and improve classifier performance. In this study, a modified ensemble classifier was developed to effectively classify PCOS using feature selection. This classification was carried out with the use of a publicly available PCOS dataset with 541 patients and 50 features. Several data cleaning, processing, and transformation techniques have been applied to the data. To extract the optimal features from the dataset, traditional machine learning models such as Decision Tree (DT), Logistic Regression (LR), K-Nearest Neighbor (KNN) and Random Forest (RF) were employed. The identified features were then evaluated, using a modified stacking classifier, to accurately classify PCOS. Various statistical metrics, including accuracy, precision, recall, mean squared error (MSE), negative predicted value (NPV), false positive rate (FPR), false negative rate (FNR), and false discovery rate (FDR) were employed to evaluate the effectiveness of the various approaches. The proposed modified stacking classifier achieved the highest accuracy of 98.61%. Experiments demonstrated that feature selection techniques enhanced the performance of all classifiers.

Currently On going Research - 2

Enhanced Depth of Anesthesia Classification Using Graph Neural Networks and EEG Features

Status: Writing on Going

Discipline: Anesthesia EEG, Medical Data, Signal preprocessing, GNN, feature extraction, feature selection,

Short Details:

This study addresses the critical need for precise monitoring of the depth of anesthesia (DoA) during surgical procedures, highlighting the limitations of conventional techniques like the Bispectral Index (BIS), which are costly and less transparent. It proposes a novel method for classifying DoA into four states: general anesthesia, profound anesthesia/burst suppression, moderate sedation, and awake/light sedation, using a dataset of 24 patient cases. The methodology employs a Graph Neural Network (GNN) after preprocessing EEG data through bandpass filtering (1–50 Hz), DC offset removal, downsampling to 128 Hz, normalization, and smoothing. A correlation-based graph is constructed to represent feature interdependencies. The GNN model achieved a classification accuracy of 91.46%. This open-source GNN paradigm offers a cost-effective and interpretable alternative to proprietary solutions, with the potential to enhance patient outcomes and accessibility in resource-limited settings.

Breast Cancer classification using Quantum Variational Neural Network

Status: Writing on Going

Discipline: Quantum Variational Neural Network, Preprocessing, feature Extraction, Classification

Page updated

Google Sites

Report abuse